H I S TO R I C A L P E R S P E C T I V E

What we knowThe cast. Alterations in three types of genes are responsible fortumorigenesis: oncogenes, tumor-suppressor genes and stabilitygenes (Tables 1 and 2). Unlike diseases such as cystic fibrosis or mus-cular dystrophy, wherein mutations in one gene can cause disease, nosingle gene defect ‘causes’ cancer. Mammalian cells have multiplesafeguards to protect them against the potentially lethal effects ofcancer gene mutations, and only when several genes are defectivedoes an invasive cancer develop. Thus it is best to think of mutatedcancer genes as contributing to, rather than causing, cancer.

Oncogenes are mutated in ways that render the gene constitu-tively active or active under conditions in which the wild-type geneis not. Oncogene activations can result from chromosomal translo-cations, from gene amplifications or from subtle intragenic muta-tions affecting crucial residues that regulate the activity of the geneproduct. For example, the most common activating mutation ofBRAF in human cancers changes a valine to a glutamate at codon599, a residue within the activation loop of the kinase domain1. Theactivation loop is normally regulated by phosphorylation at adja-cent residues (Thr598 and Ser601). This suggests that the glutamatesubstitution at codon 599 mimics a phosphate group and constitu-tively activates the enzyme even in the absence of signals that wouldnormally result in phosphorylation of the adjacent threonine orserine residues. The activated BRAF kinase then phosphorylatesdownstream targets such as extracellular signal–regulated kinase(ERK), leading to aberrant growth2(Fig. 1). A mutation in an onco-gene is analogous to a stuck accelerator in an automobile; the carstill moves forward even when the driver removes his foot from it.An activating somatic mutation in one allele of an oncogene is gen-erally sufficient to confer a selective growth advantage on the cell.

Tumor-suppressor genes are targeted in the opposite way bygenetic alterations: mutations reduce the activity of the gene

product. Such inactivations arise from missense mutations atresidues that are essential for its activity, from mutations that resultin a truncated protein, from deletions or insertions of various sizes,or from epigenetic silencing. A mutation in a tumor-suppressorgene is analogous to a dysfunctional brake in an automobile; thecar doesn’t stop even when the driver attempts to engage it. Somerecently described tumor-suppressor genes have been hypothesizedto exert a selective advantage on a cell when only one allele is inacti-vated and the other remains functional (that is, haploinsuffi-ciency)3. However, mutations in both the maternal and paternalalleles of a tumor-suppressor gene are generally required to confera selective advantage to the cell. This situation commonly arisesthrough the deletion of one allele via a gross chromosomal event—such as loss of an entire chromosome or chromosome arm—coupled with an intragenic mutation of the other allele4.

Oncogene and tumor-suppressor gene mutations all operate simi-larly at the physiologic level: they drive the neoplastic process byincreasing tumor cell number through the stimulation of cell birthor the inhibition of cell death or cell-cycle arrest. The increase can becaused by activating genes that drive the cell cycle, by inhibiting nor-mal apoptotic processes or by facilitating the provision of nutrientsthrough enhanced angiogenesis (Figs. 2–4). A third class of cancergenes, called stability genes or caretakers, promotes tumorigenesis ina completely different way when mutated. This class includes themismatch repair (MMR), nucleotide-excision repair (NER) andbase-excision repair (BER) genes responsible for repairing subtlemistakes made during normal DNA replication or induced by expo-sure to mutagens (Table 1). Other stability genes control processesinvolving large portions of chromosomes, such as those responsiblefor mitotic recombination and chromosomal segregation (for exam-ple, BRCA1, BLM and ATM; Table 1). Stability genes keep geneticalterations to a minimum, and thus when they are inactivated, muta-tions in other genes occur at a higher rate5. All genes are potentiallyaffected by the resultant increased rate of mutation, but only muta-tions in oncogenes and tumor-suppressor genes affect net cellgrowth and can thereby confer a selective growth advantage to themutant cell. As with tumor-suppressor genes, both alleles of stabilitygenes generally must be inactivated for a physiologic effect to result.

Cancer genes and the pathways theycontrolBert Vogelstein & Kenneth W Kinzler

The revolution in cancer research can be summed up in a single sentence: cancer is, in essence, a genetic disease. In the lastdecade, many important genes responsible for the genesis of various cancers have been discovered, their mutations preciselyidentified, and the pathways through which they act characterized. The purposes of this review are to highlight examples ofprogress in these areas, indicate where knowledge is scarce and point out fertile grounds for future investigation.

The authors are at the Howard Hughes Medical Institute and The Sidney KimmelComprehensive Cancer Center, The Johns Hopkins University MedicalInstitutions, Baltimore, Maryland 21231, USA.e-mail: vogelbe@welch.jhu.edu or kinzlke@jhmi.edu

Published online 30 July 2004; doi:10.1038/nm1087

NATURE MEDICINE VOLUME 10 | NUMBER 8 | AUGUST 2004 789

CELEBRATING OUR TENTH YEAR©

2004

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

emed

icin

e

H I S TO R I C A L P E R S P E C T I V E

In the analogy to autos, stability genes represent the mechanics and adefective stability gene is akin to an inept mechanic.

Mutations in these three classes of genes can occur in thegermline, resulting in hereditary predispositions to cancer (Table 1), or in single somatic cells, resulting in sporadic tumors(Tables 1 and 2). It is important to point out that a mutation isdefined as any change in the sequence of the genome. Thesechanges include those affecting single base pairs as well as thosecreating large or small deletions or insertions, amplifications ortranslocations. In the germline, the most common mutations aresubtle (point mutations or small deletions or insertions), whereasall types of mutation can be found in tumor cells. In fact, cancersrepresent one of the few disease types in which somatic mutationsoccurring after birth are pathogenic.

The first somatic mutation in an oncogene or tumor-suppressorgene that causes a clonal expansion initiates the neoplasticprocess6. Subsequent somatic mutations result in additionalrounds of clonal expansion (and thus in tumor progression)7.Indeed, the best modern definition of a neoplastic cell is one thathas clonally expanded as a result of somatic mutations6,7. Germlinemutations of these genes cause cancer predisposition, not cancerper se: people with these mutations have a ‘head start’ on the neo-plastic process, as a mutation that can contribute to cancer isalready present in every one of their cells. Such individuals there-fore often develop multiple tumors that occur at an earlier age thanin individuals whose cancer-gene mutations have all occurred

somatically4. Examples of hereditary syndromes associated withinherited mutations are listed in Table 1. In people with these syn-dromes, only a very small fraction of the total cells in an at-riskorgan become neoplastic because other (somatic) mutations arerequired to develop a clinically detectable lesion. In nearly all dom-inantly inherited syndromes caused by tumor-suppressor genesand stability genes, the first somatic mutation affects the normalcopy of the gene inherited from the unaffected parent4.Interestingly, the most common forms of hereditary cancer predis-position, leading to breast and colon cancers, are caused by inher-ited mutations of stability genes rather than tumor-suppressorgenes or oncogenes (Table 1).

The plot. As noted above, cancer-gene mutations enhance net cellgrowth. As a result of research performed over the past decade, it isclear that there are many fewer pathways than genes. This conceptis very familiar to geneticists studying yeast, flies, mice or worms—there are almost always a variety of genes that, when altered, lead tosimilar phenotypes8,9. Application of this concept to cancer hasbeen solidified by elucidation of the biochemical functions of thealtered cancer genes, either in cell culture systems, in mice or inother organisms. For example, several cancer genes directly controltransitions from a resting stage (G0 or G1) to a replicating phase(S) of the cell cycle (Fig. 2a, Rb pathway). The products of thesegenes include proteins as diverse as cdk4 (a kinase), cyclin D1(which interacts with and activates cdk4), Rb (a transcription fac-tor) and p16 (which interacts with and inhibits cdk4)10–12. Thegenes encoding Rb and p16 are tumor-suppressor genes inactivatedby mutation, whereas those encoding cdk4 and cyclin D1 are onco-genes activated by mutation. In addition to functional studies inmodel systems, detailed studies of individual tumor types have alsoprovided compelling evidence that these four genes function in asingle pathway in human cancers. Such studies have shown that themutations within this pathway obey an ‘exclusivity principle’: thatis, one and only one of the four genes noted above is generallymutated in any single tumor, exactly as predicted if the functionaleffect of each mutation was similar10–13.

Another example of the reason for focusing on pathways ratherthan individual genes has been provided by studies of the TP53tumor-suppressor gene. The p53 protein is a transcription factorthat normally inhibits cell growth and stimulates cell death wheninduced by cellular stress14–16. The most common way to disruptthe p53 pathway is through a point mutation that inactivates itscapacity to bind specifically to its cognate recognition sequence.However, there are several other ways to achieve the same effect,including amplification of the MDM2 gene and infection withDNA tumor viruses whose products (such as the E6 protein ofhuman papilloma virus) bind to p53 and functionally inactivate it(Fig. 2b, p53 pathway).

One of the most important—and most curious—discoveries ofthe 1990s was that virtually all DNA tumor viruses that causetumors in experimental animals or humans encode proteins thatinactivate both Rb and p53 (refs. 17–19). Of the hundreds of cancergenes known or remaining to be discovered, why should these twohave been singled out as targets for inactivation by all DNA tumorviruses? The answer may be that it is impossible for a tumor ofepithelial origin to form unless the p53 and Rb tumor-suppressorgene pathways have been inactivated. This conjecture is supportedby studies showing that these two pathways are altered in a largefraction of many types of cancers. We predict that most of the can-cers that now appear to be devoid of mutations in these two

790 VOLUME 10 | NUMBER 8 | AUGUST 2004 NATURE MEDICINE

ALK, EGFR, EPHB2, ERBB2, FGFR, FLT, NTRK, PDGFRB...

PDGFB

SHCABL, FES, JAK

SOS

GEF

ERK ERK

Transcriptionfactors

Transcriptionfactors

GPGtranscription

GPGtranscription

RAS:GTP

RAF

MAPKK MAPKK

RAS:GDP

NF1

STAT STATGRB2

P

P

P

P

��

�

� �� P

Figure 1 Receptor tyrosine kinase (RTK) pathway. Gene products in redboxes indicate that the corresponding genes are mutated in the germline(and generally also are somatically mutated in nonfamilial tumors; seeTable 1). Gene products in green boxes indicate that the correspondinggene has been found to be mutated only somatically (see Table 2). ‘GPG’ denotes growth-promoting-genes—that is, genes that stimulate cell proliferation or inhibit the rate of cell death or arrest. Diamonds (�)indicate protein-protein interactions. Red arrows and red T-bars indicatetranscriptional induction and repression, respectively. Small circled ‘P’,‘OH’ and ‘Ub’ represent covalently attached phosphate, hydroxyl andubiquitin groups, respectively.

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

H I S TO R I C A L P E R S P E C T I V E

NATURE MEDICINE VOLUME 10 | NUMBER 8 | AUGUST 2004 791

Table 1 Cancer predisposition genes

Gene (synonym(s))a Syndrome Hereditary pattern Second hit Pathwayb Major heredity tumor typesc

Tumor-suppressor genesAPC FAP Dominant Inactivation of WT allele APC Colon, thyroid, stomach, intestineAXIN2 Attenuated polyposis Dominant Inactivation of WT allele APC ColonCDH1 (E-cadherin) Familial gastric carcinoma Dominant Inactivation of WT allele APC StomachGPC3 Simpson-Golabi-Behmel X-linked ? APC Embryonal

syndrome

CYLD Familial cylindromatosis Dominant Inactivation of WT allele APOP Pilotrichomas

EXT1,2 Hereditary multiple Dominant Inactivation of WT allele GLI Boneexostoses

PTCH Gorlin syndrome Dominant Inactivation of WT allele GLI Skin, medulloblastomaSUFU Medulloblastoma Dominant Inactivation of WT allele GLI Skin, medulloblastoma

predisposition

FH Hereditary leiomyomatosis Dominant Inactivation of WT allele HIF1 LeiomyomasSDHB, C, D Familial paraganglioma Dominant Inactivation of WT allele HIF1 Paragangliomas,

pheochromocytomasVHL Von Hippel–Lindau syndrome Dominant Inactivation of WT allele HIF1 Kidney

TP53 (p53) Li-Fraumeni syndrome Dominant Inactivation of WT allele p53 Breast, sarcoma, adrenal, brain...

WT1 Familial Wilms tumor Dominant Inactivation of WT allele p53 Wilms’

STK11 (LKB1) Peutz-Jeghers syndrome Dominant Inactivation of WT allele PI3K Intestinal, ovarian, pancreaticPTEN Cowden syndrome Dominant Inactivation of WT allele PI3K Hamartoma, glioma, uterusTSC1, TSC2 Tuberous sclerosis Dominant Inactivation of WT allele PI3K Hamartoma, kidney

CDKN2A Familial malignant Dominant Inactivation of WT allele RB Melanoma, pancreas(p16INK4A, p14ARF) melanomaCDK4 Familial malignant Dominant ? RB Melanoma

melanomaRB1 Hereditary retinoblastoma Dominant Inactivation of WT allele RB Eye

NF1 Neurofibromatosis type 1 Dominant Inactivation of WT allele RTK Neurofibroma

BMPR1A Juvenile polyposis Dominant Inactivation of WT allele SMAD GastrointestinalMEN1 Multiple endocrine Dominant Inactivation of WT allele SMAD Parathyroid, pituitary, islet cell,

neoplasia type I carcinoidSMAD4 (DPC4) Juvenile polyposis Dominant Inactivation of WT allele SMAD Gastrointestinal

BHD Birt-Hogg-Dube syndrome Dominant Inactivation of WT allele ? Renal, hair follicleHRPT2 Hyperparathyroidism Dominant Inactivation of WT allele ? Parathyroid, jaw fibroma

Jaw-tumor syndrome.NF2 Neurofibromatosis type 2 Dominant Inactivation of WT allele ? Meningioma, acoustic neuroma

Stability genesMUTYH Attenuated polyposis Recessive ? BER Colon

ATM Ataxia telangiectasia Recessive ? CIN Leukemias, lymphomas, brainBLM Bloom syndrome Recessive ? CIN Leukemias, lymphomas, skinBRCA1, BRCA2 Hereditary breast cancer Dominant Inactivation of WT allele CIN Breast, ovaryFANCA, C, D2, E, F,G Fanconi anemia Recessive ? CIN Leukemias

NBS1 Nijmegen breakage syndrome Recessive ? CIN Lymphomas, brainRECQL4 Rothmund-Thomson syndrome Recessive ? CIN Bone, skinWRN Werner syndrome Recessive ? CIN Bone, brain

MSH2, MLH1, HNPCC Dominant Inactivation of WT allele MMR Colon, uterusMSH6, PMS2

XPA, C; ERCC2–5; Xeroderma pigmentosum Recessive ? NER SkinDDB2

OncogenesKIT Familial gastrointestinal Dominant ? RTK Gastrointestinal stromal tumors

stromal tumorsMET Hereditary papillary renal Dominant Mutant allele duplication RTK Kidney

cell carcinomaPDGFRA Familial gastrointestinal Dominant ? RTK Gastrointestinal stromal tumors

stromal tumorsRET Multiple endocrine Dominant Mutant allele duplication RTK Thyroid, parathyroid, adrenal

neoplasia type II

WT, wild type. aRepresentative genes of all the major pathways and hereditary cancer predisposition types are listed. For a complete list, see ref. 117. Approved gene symbols areprovided for each entry, with alternative names in parentheses. bIn many cases, the gene has been implicated in several pathways. The single pathway that is listed for each generepresents a ‘best guess’ (when one can be made) and for the reasons noted in the text and in the legend to Figure 9, should not be regarded as conclusive. APOP, apoptoticpathway; RTK, receptor tyrosine kinase pathway (see Fig. 1). cIn most cases, the nonfamilial tumor spectrum caused by somatic mutations of the gene includes those occurring inthe familial cases plus additional tumor types. For example, mutations of TP53 and CDKN2A are found in many more tumor types than those to which Li-Fraumeni and familialmalignant melanoma patients, respectively, are predisposed.

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

H I S TO R I C A L P E R S P E C T I V E

pathways will eventually be shown to contain them. This predictionwill become testable once more efficient methods of detectingmutations are developed and all the genes in the pathways becomeknown.

In addition to the Rb and p53 pathways, there are others thathave a role in many tumor types, including those involving adeno-matous polyposis coil (APC), glioma-associated oncogene (GLI),hypoxia-inducible transcription factor (HIF)-1, phosphoinositide3-kinase (PI3K), SMADs and receptor tyrosine kinases (RTKs)(Figs. 1 and 4–8). In each case, mutations in multiple members ofthe pathway have been found to be mutated in more than one typeof cancer, and in many cases the mutations within a single pathwayobey the exclusivity principle noted above. The discoveries of thesesignal transduction pathways represent major scientific success sto-ries of the last decade20. Their delineation has practical as well astheoretical implications. For example, theypredict that targeted therapies directedagainst a particular gene product may beactive against a tumor with a mutation ofthe targeted gene or any gene upstream ofthe target. Additionally, the fact that defectsin a relatively small number of pathwaysunderlie many different tumor types sug-gests that targeted therapeutics will beeffective against a broad range of cancers.

Tumors can be broadly classified as liquidor solid. The former includes leukemias andlymphomas, composed of neoplastic cellswhose precursors are normally mobile.Solid tumors are composed of epithelial ormesenchymal cells that normally are immo-bile. There are numerous other differencesbetween liquid and solid tumors. For exam-ple, at least three mutations seem to berequired to develop a malignant solid tumorin adults21; each of these mutations likelyalters one of the pathways described inFigure 9. In contrast, only one or two muta-tions may be required to develop a malig-nant liquid tumor6. Perhaps liquid tumorsdon’t require as many pathways to be

inactivated because their precursor cells arealready mobile and invasive, key character-istics that solid tumor cells must develop tobecome malignant. Additionally, oncogeneactivations caused by chromosome translo-cation events are the most common geneticalterations observed in liquid tumors22.Consistent chromosome translocations aremuch less common in solid tumors, whereasinactivations of tumor-suppressor genes areubiquitious23. Finally, there are severalgenes that are uniquely altered in specificsubtypes of liquid tumors, seem primarilyto affect differentiation, do not obviouslyparticipate in the pathways depicted inFigure 9 and do not occur in hereditaryform (Table 2). These molecular distinc-tions add to the cytogenetic, epidemiologicand medical evidence that liquid and solidtumors should be considered separately in

terms of their biology, behavior and pathogenesis. As solid tumorsare much more common than liquid tumors, most of the examplesprovided in this review focus on the former.

Supporting cast. That solid tumors are composed of two compart-ments, one consisting of neoplastic epithelial cells and the other ofstromal cells, was pointed out a hundred years ago24. The impor-tance of the interactions between stroma and epithelium is becom-ing increasingly recognized25,26. And four discoveries made duringthe last ten years have propelled one component of the stroma,endothelial cells, into the spotlight.

First, naturally occurring inhibitors of angiogenesis were identi-fied and shown to hinder the growth of experimental tumors27,28.Second, synthetic inhibitors of one of the major regulatory path-ways of angiogenesis (VEGF) were produced and shown to inhibit

792 VOLUME 10 | NUMBER 8 | AUGUST 2004 NATURE MEDICINE

HPV E6

p21, 14-3-3σ... PUMA, NOXA...

MDM2

Degradationp53

Ub

�

�

�

P

WT1

p53

Rb

E2F

Rb

HPV E7

TAL1

CDKN2A

p16INK4Ap14ARF

CDK Cyclin D1

CDK Cyclin

G1phase

S phase

M phaseG2phase

Notch1

BCL2 BAX

ApoptosisG1 phase S phase

TFE3

ba

Figure 2 Rb (a) and p53 (b) pathways. Symbols as in Figure 1.

TNFα

NFKβNFKβIkβ

FASL TRAIL

BH3-domain proteins

TNFR1 FAS

FADD

BID

Cyto C APAF1

Caspase 8

Caspases 3, 6, 7...

Apoptosis

Caspase 9

DR4,5

BCL2BAX

CYLD

�

�

�

� �

Figure 3 Apoptosis pathway. Symbols as in Figure 1.

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

H I S TO R I C A L P E R S P E C T I V E

tumor progression in patients with cancer29. Third, inactivation ofVHL, a classic tumor-suppressor gene, was shown to support renalcell tumor growth through the control of tumor angiogenesis30.The protein encoded by VHL is part of a ubiquitin ligase complexthat degrades HIF-1α in the presence of oxygen (Fig. 4, HIF1 pathway). In the absence of oxygen under normal circumstances orwhen VHL is mutated in tumors, the HIF-1α transcription factor isstabilized, leading to the expression of cytokines such VEGF andculminating in angiogenesis31. And finally, all oncogenes andtumor-suppressor gene pathways have been implicated in angio-genesis, either directly or indirectly, as illustrated by the centralposition of the HIF1 pathway in Figure 9. An understanding oftumor angiogenesis, blood flow, oxygenation and related issuesinvolving the tumor-host relationship is becoming essential tostudies of cancer biology as well as to the design of more effectiveforms of cancer therapy.

What we don’t knowThe opening act. The process of tumorigenesis is initiated when areplication-competent cell (stem cell or partially differentiateddescendent of a stem cell) acquires a mutation in a ‘gatekeeping’ path-way that endows it with a selective growth advantage. In some can-cers, the gatekeeper has been identified (examples are RB1, APC andNF1 in tumors of the eye, colon and nervous system, respectively). Inmost common tumors, however, the gatekeeper is not known. It isalso not known whether cancers of the lung, breast, prostate, bladderor brain can each be initiated through any one of several gatekeepingpathways or through only one. Many of the known gatekeepers wereidentified through the study of unusual families with predispositionsto specific types of cancers4. There may be other families that provideclues to the nature of the gatekeepers in the future. We predict, how-ever, that many of these remaining gatekeeping genes will only beidentified through more ‘brute-force’ approaches involving sequencedetermination of major portions of the cancer cell genome32. As gate-keeping mutations provide fundamental insights into the biology andpathogenesis of particular cancers and are of singular importance tofuture diagnostic and therapeutic strategies, further research on thistopic should be a priority.

The producers. As noted above, it appears that the cells of solidtumors must accumulate several rate-limiting mutations in cancer

genes to achieve malignant status. If these mutations had to occursimultaneously in a single cell, then the prevalence of cancer wouldbe minimal33. The current doctrine is that these mutations occurover time, with each mutation engendering a clonal expansionresulting in a large number of cells that then form a substrate forsubsequent mutations6. Are normal rates of mutation, coupledwith clonal expansions, sufficient to account for the prevalence ofcancer, or is some form of genetic instability required for a cell toundergo these multiple, sequential mutations? This issue has beenhotly debated for a long time34,35, but the last 10 years of researchhave yielded some facts that clarify the issues. First, genetic insta-bility can clearly contribute to cancer, as cogently demonstrated bythe hereditary cancer predispositions caused by defects in stabilitygenes (Table 1). Second, most cancers do not have a high mutationrate when this rate is measured at the nucleotide level. Accordingly,the number of mutations in a typical cancer cell is ∼ 1 per megabaseof DNA, similar to what would be expected in a normal cell thathad passed through as many generations and population bottle-necks36 . These observations argue against a common role fordefects in MMR, NER or BER in nonhereditary tumor types.However, there is another form of instability—chromosomal insta-bility (CIN)—that is much more commonly found in cancers, andthis is observed at the gross chromosomal rather than thenucleotide level37. Though the actual rate of chromosomal changeshas only been studied in a small number of cases, the end result ofCIN, aneuploidy, is observed in nearly all solid tumors38. At themolecular level, chromosome losses are evident as losses of het-erozygosity (LOH). An average of 25–30% of the alleles present innormal cells are lost in cancers, and it is not unusual to observelosses of over 75% of the cell’s alleles37. Classic as well as moderncytogenetic studies are fully consistent with these observations23,39.Such wholesale changes in chromosomal content can be advanta-geous for the cancer cell, allowing the efficient elimination of oneallele of a tumor-suppressor gene as well as the production of vari-ants that can rapidly adapt to changing microenvironments. Themechanisms underlying aneuploidy and chromosomal instabilityin sporadic tumors are still largely unknown, though some candi-date genes and pathways, such as those involving cell-cycle check-points, telomere crisis or centrosomes, have been proposed34,40–44.The role of telomeres in instability is particularly intriguing giventhe potential importance of telomerase in aging and the age-dependent incidence of most cancers42,45. As genetic instabilities

NATURE MEDICINE VOLUME 10 | NUMBER 8 | AUGUST 2004 793

Prolyl hydroxylase

HIF1HIF1

VHL

Elongin C

Elongin B

HIF1

�

�

�

O2

ARNT

AngiogenesisDegradation

VEGF, PDGF, EGFR

OH

Ub

MitochondriaFH

SDH

Figure 4 HIF1 pathway. Symbols as in Figure 1.

Frizzled LRP6

WNTGPC3

GSK3β

β-catenin

α-catenin

β-catenin

c-Myc

Cyclin D1

BMP4...

AXIN

E-cadherin

APC

TCF

��

��

Actin�

�

Degradation

P

��

Figure 5 APC pathway. Symbols as in Figure 1.

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

H I S TO R I C A L P E R S P E C T I V E

not only seem to be central to the neoplastic development but alsomay underlie the development of resistance to chemotherapeuticagents33,34, the identification of the molecular mechanisms respon-sible for them is an important area of study.

The ending. Although abnormalities of the cancer genes listed inTables 1 and 2 are essential contributors to cancer, most abnor-malities in these genes occur relatively early in the disease processand none are known to be specifically associated with the metastatic stage. It is this final stage—the seeding and growth ofsatellite lesions in other organs—that is ultimately responsible forthe great majority of neoplastic deaths46. Primary tumors can generally be removed through surgery but widely metastaticlesions cannot be excised and are difficult or impossible to treatwith adjuvant therapies.

Some of the biochemical processes involved in the early stages ofmetastasis, such as increased cell motility and production ofmatrix-degrading proteases, have been well studied47. However,the genetic alterations responsible for endowing cells with theseabilities have not been clearly identified. In fact, it has been suggested that metastasis is not dependent at all on new geneticabnormalities that occur after tumors have been established, andthat the propensity to metastasize is determined early in the neoplastic process rather than near its end (reviewed in ref. 48).This suggestion is not readily compatible with the evidence thatcancer is a genetic disease in which evolution occurs somatically.Just as macroevolution never stops, evolution of the cancer celldoes not stop and new variants of tumor cells with potentiallygreater capacities to invade and metastasize are always being born.This evolution is itself driven by inherent genetic instabilities, asdescribed above. However, this genetic perspective on metastasiscannot be validated—or invalidated—until a better understandingof the metastatic process is in hand.

Same actors in different roles. One might have expected that a specific mutation of a widely expressed gene would have identicalor at least similar effects in different mammalian cell types. But thisis not in general what is observed. Different effects of the samemutation are not only found in distinct cell types; differences caneven be observed in the same cell type, depending on when the

mutation occurred during the tumorigenic process. RAS genemutations provide informative examples of these complexities.

i. Cell type specificity: KRAS2 gene mutations in normal pancre-atic duct cells seem to initiate the neoplastic process, eventuallyleading to the development of pancreatic cancer49,50. The samemutations occurring in normal colonic or ovarian epithelial cellslead to self-limiting hyperplastic or borderline lesions that do notprogress to malignancy51–53.

ii. Chronology: In contrast to the effects of KRAS2 mutations in anormal colonic epithelial cell, a KRAS2 gene mutation in the samecell type that has already acquired an APC mutation results in aclonal expansion that often progresses to cancer54.

iii. Growth inhibition versus growth promotion: In many humanand experimental cancers, RAS genes seem to function as onco-genes55,56. But RAS genes can function as suppressor genes underother circumstances, inhibiting tumorigenesis after administrationof carcinogens to mice57,58.

These and similar observations on other cancer genes are consis-tent with the emerging general notion that signaling molecules playmultiple roles at multiple times, even in the same cell type (for exam-ple, see ref. 59). However, the biochemical bases for such variationsamong cancer cells are almost entirely unknown. One could arguethat oncogenes and tumor-suppressor genes are like electronic com-ponents whose effects depend on their placement within an electricalcircuit. But no such argument can be easily used to explain the celltype specificity of stability gene defects. For example, MMR genesseem to have the identical, nonredundant function in every cell typeon the planet: they limit the acquisition of specific types of muta-tions, such as those in homopolymer tracts60,61. Yet inherited mis-match repair defects lead to tumors in the colon and endometriumbut spare most other organs, including rapidly dividing, self-renewing tissues such as the small intestine and bone marrow62–64.

794 VOLUME 10 | NUMBER 8 | AUGUST 2004 NATURE MEDICINE

PTCH SMO

Fused

EXT1,2SHH

GLI

WNT, N-Myc, Cyclin D1...

SUFU

�

Figure 6 GLI pathway. Symbols as in Figure 1.

RTK PIP2 PIP2PIP3

�

�

�

�

P

P

P

P

BADP

FOXO

GPGtranscription

GPGtranslation

P

AKTP

IRS PI3K

FOXOTSC2TSC2

PAMPKAMPK

TSC1 BAD

BCL2RHEB

MTOR

4EBP1 4EBP1S6K1 S6K1

eIF4E

Apoptosis

PTENLKB1

AKT

PDK1

P

Figure 7 PI3K pathway. Symbols as in Figure 1.

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

H I S TO R I C A L P E R S P E C T I V E

It is also notable in this context that many cancer genes affect different organs when mutated in mice than when mutated inhumans8. An important practical implication of all this complexityis that it is risky to generalize the conclusions of experiments performed in any cell type or to infer the function of a gene inhuman cancer cells based on studies of the homologous genes inother organisms.

Major cast members or bit players? In addition to the genes thatare mutated in a significant portion of cancers of a given type, suchas those listed in Tables 1 and 2, there are many other genes thathave been implicated in neoplasia but not shown to be mutated.These genes have been shown to be expressed at higher or lowerlevels than expected in normal cells65,66 and are often associatedwith ‘epigenetic’ changes—that is, covalent modifications of DNAor chromatin that are preserved as the cancer cells divide67,68.Unlike genetic changes, epigenetic changes identical to those found in cancers are often found in normal cells at some stage of development.

The discovery of such genes is a growth industry now that high-throughput methods for evaluating the genes expressed in cancercells have been developed65,66,69–74. The information gained fromthese studies has proven extremely promising for the developmentof diagnostic assays, particularly for prognosis (for example, seeref. 75). But as has been shown for prostate-specific antigen (PSA),the utility of a gene for cancer diagnostics does not necessarilyreflect a causative role in the process. It will therefore be essential todetermine how genes discovered through expression-basedapproaches can be elevated from candidate status to culprit statusin human neoplasia. When genes are sometimes inactivated bymutation and in other cases inactivated by epigenetic silencing, a

cogent case can be made for their involvement in cancer. Relevantexamples include the VHL and CDKN2A tumor-suppressor genesand the hMLH1 stability gene67. In the absence of mutational evidence, functional evidence obtained from studies in vitro or innonhuman species in vivo must be used. But because of the issueswith cell type, species and chronological specificity noted above,expression and functional studies cannot currently provide the‘smoking guns’ that would definitively implicate the gene in humancancers76,77. Telomerase provides a cogent example of the difficul-ties involved in such determinations. There is abundant evidenceimplicating telomerase in processes characteristic of neoplasia,such as immortalization45,78 and genetic instability42,79, and a greatdeal is known about its biochemical properties80,81. Even so, it isunclear at present whether the expression of telomerase in cancersis abnormal, driving the neoplastic process, or simply reflects thefact that cancers are derived from stem cells that normally expressthis enzyme.

Levels of gene expression are unreliable indicators of causationbecause disturbance of any network invariably leads to a multitudeof such changes only peripherally related to the phenotype82.Without better ways to determine whether an unmutated but inter-esting candidate gene has a causal role in neoplasia, cancerresearchers will likely be spending precious time working on genesonly peripherally related to the disease they wish to study. Onechallenge for the future is therefore to develop new model systems.For liquid tumors, innovative systems have indeed been developed,such as those using immunodeficient mice reconstituted withhuman hematopoietic stem cells83. Perhaps analogous humanizedmodels can be developed to explore solid tumors in a more med-ically meaningful context than is currently possible84,85.

What makes a box-office smash? Significant gains in cancer thera-peutics have also been made in the last decade. Many of theseadvances have been incremental, but the increments add up. Thusmore potent, less toxic derivatives of classic chemotherapeuticagents have been developed, dosing and combination treatmentsfurther optimized and side effects ameliorated86. However, the

NATURE MEDICINE VOLUME 10 | NUMBER 8 | AUGUST 2004 795

SMAD2/3 SMAD2/3

SMAD4

MEN1

RUNX1

�

��

�

P

GPGtranscription

EWS

TGFβ, BMP...

Type Ireceptor

Type IIreceptor

HIF1

RTKERK

E-cadherin

WNT

TCF4

RAS:GTP

p53BAX p14ARF

BAD

GLIPI3K

Apoptosis

SMAD APC

RB

Figure 9 Overview of cancer gene pathways. The major pathways regulatingcell birth and cell death are depicted as ovals color-coded to match Figs. 1–8.The schematics in Figs. 1–8 emphasize the genes that have been shown to begenetically altered in human tumors, though many other genes participate inthese pathways. Additionally, some of the same genes appear in more thanone pathway and there is substantial ‘cross-talk’ between pathways. Selectedmediators of this cross-talk are indicated in the loops that connect thepathways. More detailed information about these pathways can be found inseveral comprehensive reviews (refs. 11,12,15,31,103–116).

Figure 8 SMAD pathway. Symbols as in Figure 1.

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

H I S TO R I C A L P E R S P E C T I V E

most dramatic therapeutic advances have come from agents targeted against proteins encoded by genes that are mutated in cancers. These include trastuzumab (Herceptin), an antibodyagainst the product of the ERBB2 gene amplified in some breastcancers87; imatinib (Gleevec), an inhibitor of tyrosine kinasesaltered in chronic myelogenous leukemia (CML)88 and gastroin-testinal stromal tumors (GISTs)89; and gefitinib (Iressa), an EGFR

kinase inhibitor used to treat lung cancers90. Though none of theseadvances have resulted in cures of many patients with advanceddisease, they can substantially improve and prolong lives. Oneimportant lesson from the use of these agents is that mutations aremore reliable indicators of a good target than is abnormal expres-sion. For example, virtually all GISTs abnormally express high lev-els of c-kit, but only those tumors with intragenic mutations of KIT

796 VOLUME 10 | NUMBER 8 | AUGUST 2004 NATURE MEDICINE

Table 2 Genes that are mutated somatically but not inherited in mutant form

Genea (synonym) Somatic mutation typeb Cancers with mutant genec Pathwayd

CTNNB1 (β-catenin) Activating codon change Colon, liver, medulloblastomas APC

BCL2 Translocation Lymphomas APOP

TNFRSF6 (FAS) Activating codon change Lymphomas, testicular germ cell tumors APOP

BAX Inactivating codon change Colon, stomach APOP

FBXW7 (CDC4) Inactivating codon change Colon, uterine, ovarian, breast CIN

GLI Amplification, translocation Brain, sarcomas GLI

HPVE6 HPV infection Cervical p53

MDM2 Amplification Sarcomas p53

NOTCH1 Translocation Leukemias p53

AKT2 Amplification Ovarian, breast PI3K

FOXO1A, 3A Translocation Rhabdomyosarcomas, leukemias PI3K

PI3KCA Activating codon change Colon, stomach, brain, breast PI3K

CCND1 (cyclin D1) Amplification, translocation Leukemias, breast RB

HPVE7 HPV infection Cervical RB

TAL1 Translocation Leukemias RB

TFE3 Translocation Kidney, sarcomas RB

ABL1 (ABL) Translocation Chronic myelogenous leukemia RTK

ALK Translocation Anaplastic large cell lymphoma RTK

BRAF Activating codon change Melanoma, colorectal, thyroid RTK

EGFR Amplification, activating codon change Glioblastomas, non–small cell lung cancers RTK

EPHB2 Inactivating codon change Prostate RTK

ERBB2 Amplification Breast, ovarian RTK

FES Activating codon change Colon RTK

FGFR1–3 Translocation Lymphomas, gastric cancers, bladder cancers RTK

FLT3, 4 Activating codon change Leukemias, angiosarcomas RTK

JAK2 Translocation Leukemias RTK

KRAS2, N-RAS Activating codon change Colorectal, pancreatic, non–small cell lung cancer RTK

NTRK1, 3 Translocation, activating codon change Thyroid, secretory breast, colon RTK

PDGFB Translocation Dermatofibrosarcomas and fibroblastomas RTK

PDGFRB Translocation Leukemias RTK

EWSR1 Translocation Ewing’s sarcomas, lymphomas, leukemias SMAD

RUNX1 Translocation Leukemias SMAD

SMAD2 Inactivating codon change Colon, breast SMAD

TGFBR1, TGFBR2 Inactivating codon change Colon, stomach, ovarian SMAD

BCL6 Translocation Lymphomas ?

EVI1 Translocation Leukemias ?

HMGA2 Translocation Lipomas ?

HOXA9, 11, 13; HOXC13, Translocation Leukemias ?

HOXD11, 13; HOX11, HOX11L2

MAP2K4 (MKK4) Inactivating codon change Pancreas, breast, colon ?

MLL Translocation, activating codon change Leukemias ?

MYC, MYCN, MYCL1 Amplification Lymphomas, neuroblastomas, small cell lung cancers ?

PTNP1, 11 Activating codon change Leukemias, colon ?

RARA Translocation Promyelocytic leukemia ?

SS18 Translocation Synovial sarcomas ?

aRepresentative genes of all the major pathways and cancer types are listed. For a complete list, see ref. 76. Approved gene symbols are provided for each entry, with alternative names in paren-theses. bActivating codon change, intragenic mutation altering one or a small number of base pairs that activates the gene product, indicating that it is an oncogene; inactivating codon change,any mutation (point mutation, small or large deletion, etc.) that inactivates the gene product, indicating that the gene is a tumor suppressor. Amplifications and translocations generally affectoncogenes, though occasional translocations disrupt a gene rather than activate it (such as has been suggested to occur with RUNX1; ref. 118). cOnly representative types of cancers are listedwhen a gene is mutated in many tumor types, Specific types of leukemias and lymphomas are listed only if a given gene is predominantly mutated in a specific subtype. dIn many cases, the genehas been implicated in several pathways. The single pathway that is listed for each gene represents a ‘best guess’ (when one can be made) and for the reasons noted in the text and in the legendto Figure 9, should not be regarded as conclusive. APOP, apoptotic pathway; RTK, receptor tyrosine kinase pathway (see Fig. 9).

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

H I S TO R I C A L P E R S P E C T I V E

respond to therapy with imatinib91. Similarly, gefitinib is efficacious only in lung cancers in which the kinase domain ofEGFR is mutated but not in lung cancers in which EGFR is simplyoverexpressed92,93.

Interestingly, we still don’t know the basis for the selectivity ofconventional chemotherapeutic agents. Why cancer cells are moresensitive to antimetabolites, alkylating agents, DNA intercalatorsand topoisomerase inhibitors than normal replicating cells is enig-matic. The explosion in research on apoptosis in the 1990s hassuggested that some tumor cells are less, rather than more, likely toundergo apoptosis after noxious stimuli94,95. Differential apop-totic proclivities therefore don’t readily explain therapeutic speci-ficities. This is consistent with studies showing that many drugsused in the clinic to treat solid tumors do not function through theinduction of apoptosis at the doses achieved in patients96. Evenwith the newer, targeted agents, the basis for selectivity is not gen-erally clear. The ERBB2 and ABL tyrosine kinases, for example, areexpressed in many normal cell types; why should targeting thesekinases lead to the demise of cancer cells but not normal cells?Concepts such as ‘cellular addiction’ offer an imaginative answerto this question, but the molecular basis for the postulated addic-tion is unknown97. Similar arguments could be made for ‘untar-geted’ drugs like Taxol; are cancer cells addicted to microtubules?It is interesting that so much work has been done on how cellsbecome resistant to conventional chemotherapeutic agents and solittle on why they are sensitive to begin with. Indeed, some studiesindicate that it is not the neoplastic cells, but rather the tumor vas-culature, that is the actual target of many conventionalchemotherapeutic agents27. Perhaps if we understood the bases forthe responsiveness of cancers to these drugs, we could devise bet-ter agents to exploit them.

The sequel. There are at least three major challenges that willoccupy most cancer researchers’ time over the next 10 years. Thefirst is the discovery of new genes that have a causal role in neoplasia, particularly those that initiate and conclude the process.The second is the delineation of the pathways through which thesegenes act and the basis for the varying actions in specific cell types.The third is the development of new ways to exploit this knowledgefor the benefit of patients.

The first two of these challenges are likely to proceed apace.Technologies for gene discovery are rapidly advancing, the humangenome will soon be in finished form, and previous analyses ofother genes provide well-traveled road maps for investigators tofollow once new genes are in hand. In fact, there are likely to be toomany genes, too many interacting proteins and too many potentialfunctions to consider, rather than too few. As noted above, cancerbiology has not kept up with cancer molecular genetics and newbiological systems are needed to separate wheat from chaff.

The third challenge, involving practical benefits to patients, willbe much more difficult to meet. There are hardly any road maps tofollow. Eradicating hundreds of billions of cancer cells from ahuman with metastatic disease is a daunting task. Each of thesecells has multiple genetic abnormalities and is capable of rapidlyevolving variants to combat any therapeutic onslaught98,99. On theother hand, many cancers respond to the somewhat crude weaponsthat have been developed to date despite their multiple geneticabnormalities. And now that cancer is no longer a ‘black box’, and isknown to be the result of alterations in a limited number of path-ways that can in principle be targeted by new generations of drugs,cautious optimism is appropriate.

But will the development of new, more specific therapeuticagents be the best way to minimize cancer morbidity and mortalityin the long-term? Most of the increased longevity that Westernsocieties enjoy today has come through better prevention ratherthan better treatment. It is clear that certain major forms of cancercan be prevented by limiting exposures to carcinogens (such assunlight and cigarette smoke in skin and lung cancers, respec-tively). Other forms of cancer can be detected early, thereby limit-ing morbidity and mortality (for example, tumors of the breast,cervix, prostate and colon). Moreover, it requires 30–40 years for atypical epithelial cell to accumulate the multiple genetic alterationsrequired to progress to metastatic disease100. This provides a hugewindow of opportunity to detect tumors at a stage when they arestill curable by conventional surgical methods. Perhaps knowledgeof the genes altered in cancer will provide the basis for new generations of diagnostic tests, employing target-specific imagingor analysis of body fluid samples ex vivo that will make early detec-tion approaches feasible101,102. Though less dramatic than cures,prevention and early detection are perhaps the most promising andfeasible means to reduce cancer deaths by the time that Nature Medicine celebrates its 20th anniversary.

ACKNOWLEDGMENTSThe authors thank many colleagues for their critical reading of the manuscriptand S. Mousses for sharing results on EphB2 before publication. Work in theauthors’ laboratories has been supported by the Ludwig Trust, the NationalColorectal Cancer Research Alliance, the Miracle Foundation, the Clayton Fundand US National Institutes of Health grants CA43460, CA57345 and CA62924.

COMPETING INTERESTS STATEMENTThe authors declare competing financial interests (see the Nature Medicinewebsite for details).

This material is part of Nature Medicine’s 10 year anniversary series. For morecontent related to these special focus issues, please see http://www.nature.com/nm/special_focus/anniversary/index.html

Published online at http://www.nature.com/naturemedicine/

1. Davies, H. et al. Mutations of the BRAF gene in human cancer. Nature 417,949–954 (2002).

2. Wan, P.T. et al. Mechanism of activation of the RAF-ERK signaling pathway byoncogenic mutations of B-RAF. Cell 116, 855–867 (2004).

3. Santarosa, M. & Ashworth, A. Haploinsufficiency for tumour suppressor genes:when you don’t need to go all the way. Biochim. Biophys. Acta 1654, 105–122(2004).

4. Knudson, A.G. Cancer genetics. Am. J. Med. Genet. 111, 96–102 (2002).5. Friedberg, E.C. DNA damage and repair. Nature 421, 436–440 (2003).6. Nowell, P.C. Tumor progression: a brief historical perspective. Semin. Cancer

Biol. 12, 261–266 (2002).7. Maley, C.C. et al. Selectively advantageous mutations and hitchhikers in neo-

plasms: p16 lesions are selected in Barrett’s esophagus. Cancer Res. 64,3414–3427 (2004).

8. Van Dyke, T. & Jacks, T. Cancer modeling in the modern era: progress and chal-lenges. Cell 108, 135–144 (2002).

9. Horvitz, H.R. Worms, life, and death. Chembiochem 4, 697–711 (2003).10. Sherr, C.J. Cancer cell cycles revisited. Cancer Res. 60, 3689–3695 (2000).11. Ortega, S., Malumbres, M. & Barbacid, M. Cyclin D-dependent kinases, INK4

inhibitors and cancer. Biochim. Biophys. Acta 1602, 73–87 (2002).12. Classon, M. & Harlow, E. The retinoblastoma tumour suppressor in develop-

ment and cancer. Nat. Rev. Cancer 2, 910–917 (2002).13. Ichimura, K. et al. Deregulation of the p14ARF/MDM2/p53 pathway is a pre-

requisite for human astrocytic gliomas with G1-S transition control gene abnor-malities. Cancer Res. 60, 417–424 (2000).

14. Vogelstein, B., Lane, D. & Levine, A.J. Surfing the p53 network. Nature 408,307–310 (2000).

15. Oren, M. Decision making by p53: life, death and cancer. Cell Death Differ. 10,431–442 (2003).

16. Prives, C. & Hall, P.A. The p53 pathway. J. Pathol. 187, 112–126 (1999).17. Klein, G. Perspectives in studies of human tumor viruses. Front. Biosci. 7,

d268–d274 (2002).18. Munger, K. & Howley, P.M. Human papillomavirus immortalization and trans-

formation functions. Virus Res. 89, 213–228 (2002).19. zur Hausen, H. Oncogenic DNA viruses. Oncogene 20, 7820–7823 (2001).

NATURE MEDICINE VOLUME 10 | NUMBER 8 | AUGUST 2004 797

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

H I S TO R I C A L P E R S P E C T I V E

20. Hunter, T. Signaling–2000 and beyond. Cell 100, 113–127 (2000).21. Komarova, N.L., Sengupta, A. & Nowak, M.A. Mutation-selection networks of

cancer initiation: tumor suppressor genes and chromosomal instability. J. Theor. Biol. 223, 433–450 (2003).

22. Rowley, J.D. The critical role of chromosome translocations in humanleukemias. Annu. Rev. Genet. 32, 495–519 (1998).

23. Mitelman, F. Recurrent chromosome aberrations in cancer. Mutat. Res. 462,247–253 (2000).

24. Verheul, H.M., Voest, E.E. & Schlingemann, R.O. Are tumours angiogenesis-dependent? J. Pathol. 202, 5–13 (2004).

25. Tlsty, T.D. & Hein, P.W. Know thy neighbor: stromal cells can contribute onco-genic signals. Curr. Opin. Genet. Dev. 11, 54–59 (2001).

26. Fata, J.E., Werb, Z. & Bissell, M.J. Regulation of mammary gland branchingmorphogenesis by the extracellular matrix and its remodeling enzymes. Breast Cancer Res. 6, 1–11 (2004).

27. Kerbel, R. & Folkman, J. Clinical translation of angiogenesis inhibitors. Nat. Rev. Cancer 2, 727–739 (2002).

28. Folkman, J. Role of angiogenesis in tumor growth and metastasis. Semin.Oncol. 29, 15–18 (2002).

29. Ferrara, N., Hillan, K.J., Gerber, H.P. & Novotny, W. Discovery and developmentof bevacizumab, an anti-VEGF antibody for treating cancer. Nat. Rev. DrugDiscov. 3, 391–400 (2004).

30. Kondo, K., Klco, J., Nakamura, E., Lechpammer, M. & Kaelin, W.G. Jr.Inhibition of HIF is necessary for tumor suppression by the von Hippel-Lindauprotein. Cancer Cell 1, 237–246 (2002).

31. Semenza, G.L. Targeting HIF-1 for cancer therapy. Nat. Rev. Cancer 3,721–732 (2003).

32. Strausberg, R.L., Simpson, A.J. & Wooster, R. Sequence-based cancergenomics: progress, lessons and opportunities. Nat. Rev. Genet. 4, 409–418(2003).

33. Loeb, L.A., Loeb, K.R. & Anderson, J.P. Multiple mutations and cancer. Proc. Natl. Acad. Sci. USA 100, 776–781 (2003).

34. Rajagopalan, H., Nowak, M.A., Vogelstein, B. & Lengauer, C. The significanceof unstable chromosomes in colorectal cancer. Nat. Rev. Cancer 3, 695–701(2003).

35. Sieber, O.M., Heinimann, K. & Tomlinson, I.P. Genomic instability—the engineof tumorigenesis? Nat. Rev. Cancer 3, 701–708 (2003).

36. Wang, T.L. et al. Prevalence of somatic alterations in the colorectal cancer cellgenome. Proc. Natl. Acad. Sci. USA 99, 3076–3080 (2002).

37. Lengauer, C., Kinzler, K.W. & Vogelstein, B. Genetic instabilities in human can-cers. Nature 396, 643–649 (1998).

38. Duesberg, P. & Li, R. Multistep carcinogenesis: a chain reaction of aneu-ploidizations. Cell Cycle 2, 202–210 (2003).

39. Albertson, D.G. & Pinkel, D. Genomic microarrays in human genetic diseaseand cancer. Hum. Mol. Genet. 12 (spec. no. 2), R145–R152 (2003).

40. Shiloh, Y. & Kastan, M.B. ATM: genome stability, neuronal development, andcancer cross paths. Adv. Cancer Res. 83, 209–254 (2001).

41. Scully, R. & Livingston, D.M. In search of the tumour-suppressor functions ofBRCA1 and BRCA2. Nature 408, 429–432 (2000).

42. Maser, R.S. & DePinho, R.A. Connecting chromosomes, crisis, and cancer.Science 297, 565–569 (2002).

43. Pihan, G. & Doxsey, S.J. Mutations and aneuploidy: co-conspirators in cancer?Cancer Cell 4, 89–94 (2003).

44. Rajagopalan, H. et al. Inactivation of hCDC4 can cause chromosomal instabil-ity. Nature 428, 77–81 (2004).

45. Shay, J.W. & Roninson, I.B. Hallmarks of senescence in carcinogenesis andcancer therapy. Oncogene 23, 2919–2933 (2004).

46. Chambers, A.F., Groom, A.C. & MacDonald, I.C. Dissemination and growth ofcancer cells in metastatic sites. Nat. Rev. Cancer 2, 563–572 (2002).

47. Fidler, I.J. Critical determinants of metastasis. Semin. Cancer Biol. 12, 89–96(2002).

48. Hunter, K.W. Host genetics and tumour metastasis. Br. J. Cancer 90, 752–755(2004).

49. Hruban, R.H., Goggins, M., Parsons, J. & Kern, S.E. Progression model forpancreatic cancer. Clin. Cancer Res. 6, 2969–2972 (2000).

50. Aguirre, A.J. et al. Activated Kras and Ink4a/Arf deficiency cooperate to pro-duce metastatic pancreatic ductal adenocarcinoma. Genes Dev. 17,3112–3126 (2003).

51. Jen, J. et al. Molecular determinants of dysplasia in colorectal lesions. CancerRes. 54, 5523–5526 (1994).

52. Pretlow, T.P. Aberrant crypt foci and K-ras mutations: earliest recognized play-ers or innocent bystanders in colon carcinogenesis? Gastroenterology 108,600–603 (1995).

53. Sieben, N.L. et al. In ovarian neoplasms, BRAF, but not KRAS, mutations arerestricted to low-grade serous tumours. J. Pathol. 202, 336–340 (2004).

54. Kinzler, K.W. & Vogelstein, B. Colorectal Tumors. in The Genetic Basis ofHuman Cancer (eds. Vogelstein, B. & Kinzler, K.W.) 565–587 (McGraw-Hill,New York, 1998).

55. Barbacid, M. ras genes. Annu. Rev. Biochem. 56, 779–827 (1987).56. Bos, J.L. ras oncogenes in human cancer: a review. Cancer Res. 49,

4682–4689 (1989).57. Zhang, Z. et al. Wildtype Kras2 can inhibit lung carcinogenesis in mice. Nat.

Genet. 29, 25–33 (2001).58. Diaz, R. et al. The N-ras proto-oncogene can suppress the malignant phenotype

in the presence or absence of its oncogene. Cancer Res. 62, 4514–4518(2002).

59. Bronner-Fraser, M. Development. Making sense of the sensory lineage. Science303, 966–968 (2004).

60. Jiricny, J. Eukaryotic mismatch repair: an update. Mutat. Res. 409, 107–121(1998).

61. Fishel, R. & Wilson, T. MutS homologs in mammalian cells. Curr. Opin. Genet.Dev. 7, 105–113 (1997).

62. Lynch, H.T. & de la Chapelle, A. Hereditary colorectal cancer. N. Engl. J. Med.348, 919–932 (2003).

63. Yamamoto, H., Imai, K. & Perucho, M. Gastrointestinal cancer of themicrosatellite mutator phenotype pathway. J. Gastroenterol. 37, 153–163(2002).

64. Honchel, R., Halling, K.C. & Thibodeau, S.N. Genomic instability in neoplasia.Semin. Cell Biol. 6, 45–52 (1995).

65. Brown, P.O. & Botstein, D. Exploring the new world of the genome with DNAmicroarrays. Nat. Genet. 21, 33–37 (1999).

66. Polyak, K. & Riggins, G.J. Gene discovery using the serial analysis of geneexpression technique: implications for cancer research. J. Clin. Oncol. 19,2948–2958 (2001).

67. Jones, P.A. & Baylin, S.B. The fundamental role of epigenetic events in cancer.Nat. Rev. Genet. 3, 415–428 (2002).

68. Feinberg, A.P. & Tycko, B. The history of cancer epigenetics. Nat. Rev. Cancer4, 143–153 (2004).

69. Collins, F.S., Green, E.D., Guttmacher, A.E. & Guyer, M.S. A vision for thefuture of genomics research. Nature 422, 835–847 (2003).

70. Schadt, E.E., Monks, S.A. & Friend, S.H. A new paradigm for drug discovery:integrating clinical, genetic, genomic and molecular phenotype data to identifydrug targets. Biochem. Soc. Trans. 31, 437–443 (2003).

71. Paddison, P.J. et al. A resource for large-scale RNA-interference-based screensin mammals. Nature 428, 427–431 (2004).

72. Berns, K. et al. A large-scale RNAi screen in human cells identifies new com-ponents of the p53 pathway. Nature 428, 431–437 (2004).

73. Rosenblatt, K.P. et al. Serum proteomics in cancer diagnosis and management.Annu. Rev. Med. 55, 97–112 (2004).

74. Luo, J., Isaacs, W.B., Trent, J.M. & Duggan, D.J. Looking beyond morphology:cancer gene expression profiling using DNA microarrays. Cancer Invest. 21,937–949 (2003).

75. Ma, X.J. et al. A two-gene expression ratio predicts clinical outcome in breastcancer patients treated with tamoxifen. Cancer Cell 5, 607–616 (2004).

76. Futreal, P.A. et al. A census of human cancer genes. Nat. Rev. Cancer 4,177–183 (2004).

77. Masayesva, B.G. et al. Gene expression alterations over large chromosomalregions in cancers include multiple genes unrelated to malignant progression.Proc. Natl. Acad. Sci. USA 101, 8715–8720 (2004).

78. Stewart, S.A. & Weinberg, R.A. Senescence: does it all happen at the ends?Oncogene 21, 627–630 (2002).

79. Feldser, D.M., Hackett, J.A. & Greider, C.W. Telomere dysfunction and the ini-tiation of genome instability. Nat. Rev. Cancer 3, 623–627 (2003).

80. Chan, S.R. & Blackburn, E.H. Telomeres and telomerase. Phil. Trans. R. Soc.Lond. B 359, 109–121 (2004).

81. Cech, T.R. Beginning to understand the end of the chromosome. Cell 116,273–279 (2004).

82. Miklos, G.L. & Maleszka, R. Microarray reality checks in the context of a com-plex disease. Nat. Biotechnol. 22, 615–621 (2004).

83. Hope, K.J., Jin, L. & Dick, J.E. Human acute myeloid leukemia stem cells.Arch. Med. Res. 34, 507–514 (2003).

84. Berking, C. & Herlyn, M. Human skin reconstruct models: a new application forstudies of melanocyte and melanoma biology. Histol. Histopathol. 16,669–674 (2001).

85. Kuperwasser, C. et al. Reconstruction of functionally normal and malignanthuman breast tissues in mice. Proc. Natl Acad. Sci. USA 101, 4966–4971(2004).

86. Frei, E.I. & Eder, J.P. Principles of dose, schedule, and combination Therapy.in Cancer Medicine (eds. Kufe, D.W. et al.) 669–677 (B.C. Decker, Inc.,Hamilton, Ontario, 2003).

87. Pegram, M.D., Konecny, G. & Slamon, D.J. The molecular and cellular biologyof HER2/neu gene amplification/overexpression and the clinical developmentof herceptin (trastuzumab) therapy for breast cancer. Cancer Treat. Res. 103,57–75 (2000).

88. Druker, B.J. et al. Chronic myelogenous leukemia. in Hematology 2001(American Society of Hematology Education Program) 87–112 (AmericanSociety of Hematology, 2001).

89. Mechtersheimer, G. et al. Gastrointestinal stromal tumours and their responseto treatment with the tyrosine kinase inhibitor imatinib. Virchows Arch. 444,108–118 (2004).

90. Langer, C.J. Emerging role of epidermal growth factor receptor inhibition intherapy for advanced malignancy: focus on NSCLC. Int. J. Radiat. Oncol. Biol.Phys. 58, 991–1002 (2004).

91. Duensing, A., Heinrich, M.C., Fletcher, C.D. & Fletcher, J.A. Biology of gas-trointestinal stromal tumors: KIT mutations and beyond. Cancer Invest. 22,106–116 (2004).

92. Paez, J.G. et al. EGFR mutations in lung cancer: correlation with clinicalresponse to gefitinib therapy. Science 304, 1497–1500 (2004).

798 VOLUME 10 | NUMBER 8 | AUGUST 2004 NATURE MEDICINE

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine

H I S TO R I C A L P E R S P E C T I V E

93. Lynch, T.J. et al. Activating mutations in the epidermal growth factor receptorunderlying responsiveness of non-small-cell lung cancer to gefitinib. N. Engl.J. Med. 350, 2129–2139 (2004).

94. Schmitt, C.A. & Lowe, S.W. Apoptosis and therapy. J. Pathol. 187, 127–137(1999).

95. Danial, N.N. & Korsmeyer, S.J. Cell death: critical control points. Cell 116,205–219 (2004).

96. Brown, J.M. & Wouters, B.G. Apoptosis: mediator or mode of cell killing byanticancer agents? Drug Resist. Updat. 4, 135–136 (2001).

97. Weinstein, I.B. et al. Disorders in cell circuitry associated with multistage car-cinogenesis: exploitable targets for cancer prevention and therapy. Clin. CancerRes. 3, 2696–2702 (1997).

98. Nygren, P. & Larsson, R. Overview of the clinical efficacy of investigationalanticancer drugs. J. Intern. Med. 253, 46–75 (2003).

99. Shih, L.Y. et al. Heterogeneous patterns of FLT3 Asp(835) mutations inrelapsed de novo acute myeloid leukemia: a comparative analysis of 120paired diagnostic and relapse bone marrow samples. Clin. Cancer Res. 10,1326–1332 (2004).

100. Kinzler, K.W. & Vogelstein, B. Lessons from hereditary colon cancer. Cell 87,159–170 (1996).

101. Weissleder, R. & Ntziachristos, V. Shedding light onto live molecular targets.Nat. Med. 9, 123–128 (2003).

102. Sidransky, D. Emerging molecular markers of cancer. Nat. Rev. Cancer 2,210–219 (2002).

103. Gschwind, A., Fischer, O.M. & Ullrich, A. The discovery of receptor tyrosinekinases: targets for cancer therapy. Nat. Rev. Cancer 4, 361–370 (2004).

104. Downward, J. Targeting RAS signalling pathways in cancer therapy. Nat. Rev.Cancer 3, 11–22 (2003).

105. Malumbres, M. & Barbacid, M. To cycle or not to cycle: a critical decision in

cancer. Nat. Rev. Cancer 1, 222–231 (2001).106. Giles, R.H., van Es, J.H. & Clevers, H. Caught up in a Wnt storm: Wnt signaling

in cancer. Biochim. Biophys. Acta 1653, 1–24 (2003).107. Cantley, L.C. The phosphoinositide 3-kinase pathway. Science 296,

1655–1657 (2002).108. Shi, Y. & Massague, J. Mechanisms of TGF-β signaling from cell membrane to

the nucleus. Cell 113, 685–700 (2003).109. Ruiz i Altaba. A., Stecca, B. & Sanchez, P. Hedgehog–Gli signaling in brain

tumors: stem cells and paradevelopmental programs in cancer. Cancer Lett.204, 145–157 (2004).

110. Adams, J.M. Ways of dying: multiple pathways to apoptosis. Genes Dev. 17,2481–2495 (2003).

111. Blagosklonny, M.V. & Pardee, A.B. The restriction point of the cell cycle. CellCycle 1, 103–110 (2002).

112. Plas, D.R. & Thompson, C.B. Cell metabolism in the regulation of programmedcell death. Trends Endocrinol. Metab. 13, 75–78 (2002).

113. Green, D.R. & Evan, G.I. A matter of life and death. Cancer Cell 1, 19–30(2002).

114. Eng, C., Kiuru, M., Fernandez, M.J. & Aaltonen, L.A. A role for mitochondrialenzymes in inherited neoplasia and beyond. Nat. Rev. Cancer 3, 193–202(2003).

115. Lum, L. & Beachy, P.A. The Hedgehog response network: sensors, switches,and routers. Science 304, 1755–1759 (2004).

116. Brivanlou, A.H. & Darnell, J.E. Jr. Signal transduction and the control of geneexpression. Science 295, 813–818 (2002).

117. Vogelstein, B. & Kinzler, K.W. The Genetic Basis of Human Cancer (McGraw-Hill, Toronto, 2002).

118. Cameron, E.R. & Neil, J.C. The Runx genes: lineage-specific oncogenes andtumor suppressors. Oncogene 23, 4308–4314 (2004).

NATURE MEDICINE VOLUME 10 | NUMBER 8 | AUGUST 2004 799

©20

04 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

urem

edic

ine